OpenAI Enhances ChatGPT’s Image Creation Capabilities

OpenAI has introduced a new image generation AI model on Tuesday, titled ChatGPT Images 2.0. This model is capable of producing multiple images from a single prompt, such as an entire study guide, as well as generating text in various languages, including Chinese and Hindi. This release is globally accessible for both ChatGPT and Codex users, with an upgraded version available for subscribers.

When any significant AI company unveils a new image model, it tends to rekindle interest and increase usage, particularly if social media users start a meme trend by transforming their own images. Last year, Google’s launch of the Nano Banana model marked a pivotal moment for the company, especially as users began sharing hyperrealistic figurines of themselves online. Earlier this year, ChatGPT Images created a buzz on social media as users shared AI-generated caricatures.

What’s Different?

The new model can leverage ChatGPT’s “reasoning” capabilities, enabling Images 2.0 to search the internet for up-to-date information and produce multiple images simultaneously. Essentially, the bot can employ additional steps to create more comprehensive outputs from a single prompt. Images 2.0 also boasts a more recent knowledge cutoff of December 2025.

This enhancement means that the outputs from the new model are more nuanced. For instance, I created an infographic detailing San Francisco’s weather forecast for the following day, along with suggested activities. The image generated by ChatGPT provided accurate weather information for the rainy day, as well as lifelike drawings of the Ferry Building, Castro Theater, Painted Ladies houses, and Transamerica Pyramid.

Moreover, Images 2.0 offers greater customization for users seeking unique aspect ratios in their image outputs. The new model can generate images that vary from 3:1 wide to 1:3 tall, and users have the option to modify the image size as part of their prompt to the AI tool.

First Impressions

After spending a few hours experimenting with the new model, I found myself quite impressed with its text rendering capabilities, at least in English. Not long ago, image outputs with text from any major models frequently produced malformed characters or words with unintended extra letters. ChatGPT faced challenges in accurately labeling images two years ago, making the cleaner, more sophisticated outputs from Images 2.0 indicate significant progress. Google has also dedicated efforts to enhancing text-included image outputs in its recent iterations of Nano Banana.

AI-GENERATED BY REECE ROGERS